Rofi spread by anton-seaice · Pull Request #13 · ACCESS-NRI/CMEPS

anton-seaice · 2026-03-01T23:56:56Z

This change allows for spreading of the iceberg melt field according to a climatological pattern.

A new configuration parameter is added, rof2ocn_ice_spread - which is the filename for a netcdf file with a 12-monthly climatological spreading pattern. If this parameter is not set - there is no change to existing behaviour (see repro test)

When a file is provided, the incoming Forr_rofi field is spread according to the pattern supplied in the climatology of spreading.

e.g., for January:

and for June:

Water volume is maintained between that coming from drof and ficeberg in MOM:

See example OM3 100km configuration:

ACCESS-NRI/access-om3-configs@3169a38

anton-seaice · 2026-03-04T04:51:04Z

mediator timestamps are a bit strange, hence the minor difference in mediator vs mom runoff plots:

aekiss · 2026-03-10T00:07:03Z

Thanks @anton-seaice - do you think this is ready for the next 25km IAF test run?

anton-seaice · 2026-03-10T00:09:00Z

Ready from my perspective.

Do we want to seperate the impact of changing the land mask in the poles to support b-grid advection from this change ?
Needs review and deployment - would take a few days at least

anton-seaice · 2026-03-13T03:48:34Z

There isn't much flexibility in the mediator io. I modified this so that it can read 3d variable (x,y,time) from a netcdf . For whatever reason, 3d information in restart files is saved in separate 2d variables so the functionality didn't exisit.

anton-seaice · 2026-03-18T00:38:40Z

I did some cleanup, but have finished that for now

dougiesquire

Nice work @anton-seaice. Sorry for taking so long to get to this.

My main worry is the lack of time interpolation and the resulting step changes in the pattern between months. But, as we've discussed, we probably can't address that here.

dougiesquire · 2026-04-15T01:55:08Z

    !-------------------------------------------------------------------------------
 #endif

+   rc = ESMF_SUCCESS


Why is this needed? Is this just convention (e.g. in case someone adds an early return path or something)?

dougiesquire · 2026-04-15T03:01:51Z

+        runoff_flux => rof2ocn_spread(:,month)
+        ! calculate sum of spreading 
+        local_sum = 0.0_r8
+        do i = 1, size(runoff_flux)
+          if (lats(i) < 0.0_r8) then
+            local_sum(1) = local_sum(1) + areas(i) * runoff_flux(i)
+          else
+            local_sum(2) = local_sum(2) + areas(i) * runoff_flux(i)
+          end if
+        end do
+
+        call ESMF_VMAllreduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &
+            reduceflag=ESMF_REDUCE_SUM, rc=rc)
+        if (ChkErr(rc,__LINE__,u_FILE_u)) return
+
+        ! adjust correction so that it's sums to 1 in each hemisphere
+        do i = 1, size(runoff_flux)
+          if (lats(i) < 0.0_r8) then
+            runoff_flux(i) = runoff_flux(i) / global_sum(1)
+          else
+            runoff_flux(i) = runoff_flux(i) / global_sum(2)
+          end if
+        end do
+
+        rof2ocn_spread(:,month) = runoff_flux


runoff_flux is a pointer, so the final assignment here is not necessary.

Personal preference, but I think the use of a pointer here just adds indirection.

Suggested change

runoff_flux => rof2ocn_spread(:,month)

! calculate sum of spreading

local_sum = 0.0_r8

do i = 1, size(runoff_flux)

if (lats(i) < 0.0_r8) then

local_sum(1) = local_sum(1) + areas(i) * runoff_flux(i)

else

local_sum(2) = local_sum(2) + areas(i) * runoff_flux(i)

end if

end do

call ESMF_VMAllreduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &

reduceflag=ESMF_REDUCE_SUM, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

! adjust correction so that it's sums to 1 in each hemisphere

do i = 1, size(runoff_flux)

if (lats(i) < 0.0_r8) then

runoff_flux(i) = runoff_flux(i) / global_sum(1)

else

runoff_flux(i) = runoff_flux(i) / global_sum(2)

end if

end do

rof2ocn_spread(:,month) = runoff_flux

! calculate sum of spreading

local_sum = 0.0_r8

do i = 1, size(areas)

if (lats(i) < 0.0_r8) then

local_sum(1) = local_sum(1) + areas(i) * rof2ocn_spread(i,month)

else

local_sum(2) = local_sum(2) + areas(i) * rof2ocn_spread(i,month)

end if

end do

call ESMF_VMAllreduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &

reduceflag=ESMF_REDUCE_SUM, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

! adjust correction so that it's sums to 1 in each hemisphere

do i = 1, size(areas)

if (lats(i) < 0.0_r8) then

rof2ocn_spread(i,month) = rof2ocn_spread(i,month) / global_sum(1)

else

rof2ocn_spread(i,month) = rof2ocn_spread(i,month) / global_sum(2)

end if

end do

dougiesquire · 2026-04-15T03:21:57Z

+    if (maintask .and. dbug_flag > dbug_threshold) then
+
+      ! calculate the new global sum (after correction), should be equal to 0
+      local_sum = 0.0_r8
+      do n = 1, size(runoff_flux)
+        if (lats(n) < 0.0_r8) then
+          local_sum(1) = local_sum(1) + areas(n) * runoff_flux(n)
+        else
+          local_sum(2) = local_sum(2) + areas(n) * runoff_flux(n)
+        end if
+      end do
+
+      call ESMF_GridCompGet(gcomp, vm=vm, rc=rc)
+      if (ChkErr(rc,__LINE__,u_FILE_u)) return
+      call ESMF_VMReduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &
+          reduceflag=ESMF_REDUCE_SUM, rootPet = 0, rc=rc)
+      if (ChkErr(rc,__LINE__,u_FILE_u)) return
+      write(logunit,'(a)') subname//' Before correction: '//trim(field_name)
+      write(logunit,'(a,e27.17)') subname//' global_sh = ', global_sum(1)
+      write(logunit,'(a,e27.17)') subname//' global_nh = ', global_sum(2)
+    end if


Should this really all be inside if (maintask)??

Also, is maintask is always PE 0? If it isn't, you may need to use ESMF_VMAllreduce. If it is, you could do:

Suggested change

if (maintask .and. dbug_flag > dbug_threshold) then

! calculate the new global sum (after correction), should be equal to 0

local_sum = 0.0_r8

do n = 1, size(runoff_flux)

if (lats(n) < 0.0_r8) then

local_sum(1) = local_sum(1) + areas(n) * runoff_flux(n)

else

local_sum(2) = local_sum(2) + areas(n) * runoff_flux(n)

end if

end do

call ESMF_GridCompGet(gcomp, vm=vm, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

call ESMF_VMReduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &

reduceflag=ESMF_REDUCE_SUM, rootPet = 0, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

write(logunit,'(a)') subname//' Before correction: '//trim(field_name)

write(logunit,'(a,e27.17)') subname//' global_sh = ', global_sum(1)

write(logunit,'(a,e27.17)') subname//' global_nh = ', global_sum(2)

end if

if (dbug_flag > dbug_threshold) then

! calculate the new global sum (after correction), difference should be equal to 0

local_sum = 0.0_r8

do n = 1, size(runoff_flux)

if (lats(n) < 0.0_r8) then

local_sum(1) = local_sum(1) + areas(n) * runoff_flux(n)

else

local_sum(2) = local_sum(2) + areas(n) * runoff_flux(n)

end if

end do

call ESMF_GridCompGet(gcomp, vm=vm, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

call ESMF_VMReduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &

reduceflag=ESMF_REDUCE_SUM, rootPet=0, rc=rc)

if (ChkErr(rc,__LINE__,u_FILE_u)) return

if (maintask) then

write(logunit,'(a)') subname//' After correction: '//trim(field_name)

write(logunit,'(a,e27.17)') subname//' global_sh = ', global_sum(1)

write(logunit,'(a,e27.17)') subname//' global_nh = ', global_sum(2)

end if

end if

dougiesquire · 2026-04-15T03:22:19Z

+      call ESMF_VMReduce(vm, senddata=local_sum, recvdata=global_sum, count=2, &
+          reduceflag=ESMF_REDUCE_SUM, rootPet = 0, rc=rc)
+      if (ChkErr(rc,__LINE__,u_FILE_u)) return
+      write(logunit,'(a)') subname//' Before correction: '//trim(field_name)


Suggested change

write(logunit,'(a)') subname//' Before correction: '//trim(field_name)

write(logunit,'(a)') subname//' After correction: '//trim(field_name)

dougiesquire · 2026-04-15T03:22:40Z

+
+    if (maintask .and. dbug_flag > dbug_threshold) then
+
+      ! calculate the new global sum (after correction), should be equal to 0


Suggested change

! calculate the new global sum (after correction), should be equal to 0

! calculate the new global sum (after correction), difference should be equal to 0

dougiesquire · 2026-04-15T04:11:31Z

       rcode = pio_inq_dimlen(pioid, dimid(1), lnx)
+       if (rcode /= pio_noerr) then
+          ierr = pio_strerror(rcode, tmpstr)
+          call shr_sys_abort(trim(subname)//' ERROR: '//trim(tmpstr), rc=rc)


How come you call shr_sys_abort here, but shr_log_error for lny and lni below?

dougiesquire · 2026-04-15T04:20:29Z

+  subroutine med_io_read_init_iodesc(field, name1, pioid, iodesc, ungridded_nc, rc)

    use ESMF , only : ESMF_LogWrite, ESMF_LOGMSG_INFO, ESMF_SUCCESS
    use ESMF , only : ESMF_FieldBundleIsCreated, ESMF_FieldBundle, ESMF_Mesh, ESMF_DistGrid


ESMF_FieldBundle is no longer used (same for a number of other imports, but this one you removed)

Suggested change

use ESMF , only : ESMF_FieldBundleIsCreated, ESMF_FieldBundle, ESMF_Mesh, ESMF_DistGrid

use ESMF , only : ESMF_FieldBundleIsCreated, ESMF_Mesh, ESMF_DistGrid

dougiesquire · 2026-04-15T04:23:02Z

+    integer                       :: rcode, ierr
    integer                       :: nf
-    integer                       :: k,n,l
+    integer                       :: k,n,l,m


Unused variable

Suggested change

integer :: k,n,l,m

integer :: k,n,l

dougiesquire · 2026-04-15T04:23:30Z

    character(len=*)              ,optional ,intent(in)  :: pre      ! prefix to variable name
+    logical                       ,optional ,intent(in)  :: ungridded_nc
+         ! if true : ungridded dim in fields is dimension in netcdf file, 
+         ! if false: ungridded_dim is provided in seperate variables with the index appended


Suggested change

! if false: ungridded_dim is provided in seperate variables with the index appended

! if false: ungridded_dim is provided in separate variables with the index appended

dougiesquire · 2026-04-15T04:45:12Z

+                call ESMF_LogWrite(trim(subname)//trim(tmpstr), ESMF_LOGMSG_INFO, rc=rc)
+                fldptr2 = 0.0_r8


How come we don't abort here?

anton-seaice added 3 commits February 26, 2026 13:31

some work

bba7766

working

2efa4ee

remove some debugging

df6a116

anton-seaice self-assigned this Mar 1, 2026

anton-seaice added 3 commits March 2, 2026 10:59

tidyup

f42d5ed

fix syntax

514deb0

corrections / neatness

3fb4f00

anton-seaice marked this pull request as ready for review March 4, 2026 03:05

anton-seaice requested a review from dougiesquire March 4, 2026 03:06

anton-seaice commented Mar 13, 2026

View reviewed changes

Comment thread mediator/med_io_mod.F90 Outdated

anton-seaice and others added 2 commits March 18, 2026 11:36

Some cleanup

988c698

Apply suggestion from @anton-seaice

0153923

dougiesquire requested changes Apr 15, 2026

View reviewed changes

	write(logunit,'(a)') subname//' Before correction: '//trim(field_name)
	write(logunit,'(a)') subname//' After correction: '//trim(field_name)


		if (maintask .and. dbug_flag > dbug_threshold) then

		! calculate the new global sum (after correction), should be equal to 0

	use ESMF , only : ESMF_FieldBundleIsCreated, ESMF_FieldBundle, ESMF_Mesh, ESMF_DistGrid
	use ESMF , only : ESMF_FieldBundleIsCreated, ESMF_Mesh, ESMF_DistGrid

	! if false: ungridded_dim is provided in seperate variables with the index appended
	! if false: ungridded_dim is provided in separate variables with the index appended

		call ESMF_LogWrite(trim(subname)//trim(tmpstr), ESMF_LOGMSG_INFO, rc=rc)
		fldptr2 = 0.0_r8

Conversation

anton-seaice commented Mar 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

anton-seaice commented Mar 4, 2026

Uh oh!

aekiss commented Mar 10, 2026

Uh oh!

anton-seaice commented Mar 10, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

anton-seaice commented Mar 18, 2026

Uh oh!

dougiesquire left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

anton-seaice commented Mar 1, 2026 •

edited

Loading